整数反直觉测试问题：R语言整数判断函数对45×1.4返回FALSE

Great question—this is a classic floating-point precision quirk that catches even experienced R users off guard! Let’s break down exactly why your is.wholeNumber function fails with 45*1.4, and how to fix it.

What’s Really Happening with 45*1.4 #

At first glance, 45 * 1.4 should equal 63 exactly. But binary floating-point numbers (which R uses for numeric values) can’t represent every decimal fraction perfectly. Let’s peek at the actual value stored:

print(45 * 1.4, digits = 20)
# Output: 62.99999999999999289457

That tiny trailing decimal is why your function returns FALSE: floor(45*1.4) becomes 62, and 62.99999999999999... isn’t equal to 62.

Why Your Original Function Fails #

Your function is.wholeNumber <- function(x) x == floor(x) relies on exact equality, which is risky with floating-point numbers. Even calculations that should result in an integer can have tiny precision errors due to how decimals are converted to binary.

Fixing the Integer Check #

Instead of checking for exact equality, we should check if the value is "close enough" to an integer, within a small tolerance (called epsilon). Here are a few robust approaches:

1. Use a Tolerance Threshold #

Compare the absolute difference between the value and its rounded version to a small number (like 1e-10, which is way smaller than any meaningful decimal you’d work with):

is.wholeNumber <- function(x) {
  abs(x - round(x)) < 1e-10
}

# Test it out
is.wholeNumber(45*1.4)  # Returns TRUE
is.wholeNumber(63)      # Returns TRUE
is.wholeNumber(63.1)    # Returns FALSE

2. Use all.equal() for Flexible Comparison #

R’s built-in all.equal() function accounts for floating-point precision by default, so you can use it to check if the value matches its rounded integer:

is.wholeNumber <- function(x) {
  isTRUE(all.equal(x, round(x)))
}

Wrapping it in isTRUE() ensures you get a clean boolean output instead of a character message about differences.

3. Use dplyr::near() (If You’re Using the Tidyverse) #

The near() function from dplyr is designed specifically for this kind of floating-point comparison:

library(dplyr)
is.wholeNumber <- function(x) {
  near(x, round(x))
}

Key Takeaway #

Never rely on exact equality (==) when working with floating-point numbers. Always use a tolerance-based check to account for tiny precision errors that come from decimal-to-binary conversion.

内容的提问来源于stack exchange，提问作者owen88